Research On QoS Property Modeling And Verification Of Web Service Composition

With the rapid development of Internet, the traditional distributed computing technology cannot meet the users’ demands. Under this circumstance, Web service technology comes forth and is widely accepted as a paradigm of service-oriented computing. Because of its encapsulation, loose coupling and interoperability, Web services have vast potential for future development.With the development of Web services, more and more enterprises package its business processes as standard Web services and release the Web services to the Internet, which causes the great increase of Web services on the Internet. Among the published Web services, there exist many Web services which have the same function but different QoS. By using the increasing Web services, there can be thousands of combination plans which have the same function but different QoS. It is a key problem of Web services to decide whether the combination plan meets the user’s requirement.At present, the research on Web services composition with QoS constrains is still on its early stage of study. This paper carries out exploratory research on Web services composition with QoS constrains with the purpose of improving the precision of validation. The paper’s main work includes three aspects; first, by analyzing the merits and demerits of existing QoS models, we present a QoS model based on ontology. This model is based on OWL-S, it extends the Process Model. The model has good scalable and it can meet the requirements of the description of service quality. Second, we present an approach to convert Web service composition models described by extended OWLS into Markov models, and then we convert Markov models into the input language of probabilistic model checker PRISM. Algorithms are given to realize the transformation method. At last, by analyzing the merits and demerits of existing algorithms, KSP and XBF algorithms, we present a new algorithm called XKSP which combines the merits of KSP and XBF algorithms. XKSP algorithm uses KSP algorithm to produce a diagnostic path, and then store this path into a diagram which comes from XBF algorithm. XKSP algorithm can avoid the demerit of KSP algorithm, that is, the KSP algorithm needs to enumerate large number of diagnostic paths, and it can avoid the demerit of XBF algorithm, that is, the XBF algorithm may produce suboptimal diagnostic paths.